基于GPU的声场FDTD算法研究与实现

The Research and Simulation of Acoustic FDTD Method based on GPU

时域有限差分(FDTD)方法在声波时域仿真中是一项重要的技术。但巨大的运算量是制约它应用的一个重要因素。在这篇论文中,介绍基于FDTD方法使用图形处理单元(GPU)对声压在时域上进行数据估算。与计算机相比,GPU的应用大大缩短了声场FDTD仿真的运行时间,使得声场时域仿真的研究变得可行,对以前因需要大规模计算而不能实践的声场其他问题的研究开辟了新的途径。

The finite difference time domain （FdTd） method is an important technique used in time-domain acoustic simulation. However, the huge computing capacity has been a limiting factor for its applications. In this paper, we will present a graphics processing units （gPu）-based parallel FdTd framework and its successful application to the estimation of time-series sound pressure data. Compared with the computer, the runtime of acoustic ifeld FDTD simulation on gPu is dramatically decreased. The acceleration in runtime will pave the way for other studies of large-scale computational acoustic problems which were previously impractical.